Re: [PATCH v2 32/58] objtool/klp: Add test for EXPORT_SYMBOL_FOR_MODULES references

On Sun, Sep 13, 2026 at 11:45 PM [off-list ref] wrote:

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] The test script claims to cover commit d8a2860b4a36 but fails to exercise its logic because it patches vmlinux instead of a module.
--

commit 0018f6f310cd6d22138d9d2b74d24fb1f9be4ab3
Author: Song Liu [off-list ref]

objtool/klp: Add test for EXPORT_SYMBOL_FOR_MODULES references

This commit adds a test script to verify klp relocations for symbols
exported with EXPORT_SYMBOL_FOR_MODULES(). The test aims to cover the
behavior of previous fixes related to klp relocation section naming.

quoted

diff --git a/tools/objtool/tests/generic/test-export-symbol-for-modules.sh b/tools/objtool/tests/generic/test-export-symbol-for-modules.sh
new file mode 100755
index 0000000000000..7e7bdde6a7aca
--- /dev/null
+++ b/tools/objtool/tests/generic/test-export-symbol-for-modules.sh

@@ -0,0 +1,39 @@
+#!/bin/bash

[ ... ]

quoted

+. "$(dirname "$0")/../lib.sh"
+
+setup
+build_pair cross_module.c

[Severity: Medium]
Does this test actually exercise the module-patching case fixed in
d8a2860b4a36?

By calling build_pair without a -DMODNAME flag, the test fixture defaults to
patching vmlinux. Because the patched object is vmlinux, the
__klp_relocs.vmlinux section is generated regardless of the logic introduced
in d8a2860b4a36.

That commit forces the section to vmlinux when the target is a vmlinux
symbol, overriding the patched module's name. As a result of patching
vmlinux directly here, the test appears to pass even without the fix it
claims to cover.

Should this test pass a module name to build_pair to properly test the module
relocation behavior?

Good catch! Fixed in v3.

Song